分布式网络拥塞控制算法稳定性研究

Stability study of distributed network congestion control algorithm

为解决由于网络阻塞而造成的分组丢弃和时延,基于优化理论框架,建立了原始—对偶模型,可用于描述一类拥塞控制算法。对该模型平衡点存在的条件和系统的稳定性进行了分析,与传统的网络控制算法相比,该算法考虑到边界条件的限制,得出了系统存在唯一平衡点的条件。通过TCP/E-RED网络对该理论进行了验证,结果表明,对于一般网络结构,该模型不具有往返时延,系统具有极高的稳定性,提高了网络性能。该方法用于避免由于网络阻塞而造成的分组丢弃和时延是可行的、有效的。

In order to solve the problem of packet dropping and delay caused by network congestion, based on the framework of theoretical optimization, this paper established a pair of original-dual model, which could be used to describe a class of congestion control algorithms. It analyzed the stability of the system conditions and the presence of the equilibrium point. Compared with the conventional network control algorithm, the limits of the boundary condition was taken into account, the algorithm obtained conditions of the system＇ s unique equilibrium point. Validating the theory via TCP/E-RED network, the results show that, for the general structure of the network, the model does not have the round-trip delay, system owns high stability, network performance is improved. The method is feasible and effective when it was used to avoid packet dropping and delay caused by network congestion.